Elasticsearch 2.1 node ID changed after restart


(Radu Stefanache) #1

Hello everyone,

I have a cluster with 3 nodes, each node has the name set in the elasticsearch.yml
Today I've upgraded to ES 2.1 from 2.0 and found an interesting behaviour .
Whenever I restart a node, it comes back with a totally different ID .

Example :
[2015-11-25 11:37:14,356][INFO][cluster.service][node1] removed {{node2}{m6A6gifIT527eqLasHxP_g}
[2015-11-25 11:37:22,328][INFO][cluster.service][node1] added {{node2}{nsoiqwmVTu6EC-dQnEoLaQ}

If you are using Marvel, you will end up with a stale node, based on the old id .
Did anyone else bumped into this ?
In 2.0, the nodes were coming back with the same id .


(Christoph) #2

Hi,

the change of the node id is actually expected behaviour after restart. The id will be randomly generated unless you specify a specific seed ("discovery.id.seed") in your settings. Unless you did that in your 2.0 installation, I'm wondering how you would get the same id there.


(Radu Stefanache) #3

I didn't used the discovery.id.seed .
The only thing that I am not using is the multicast discovery mode so for me :

discovery.zen.ping.multicast.enabled: false
discovery.zen.ping.unicast.hosts: ["node2", "node3"]

It's weird but all looks fine now, after a while I see that Marvel is refreshing the nodes and the ones with the old id disappear .

Thanks for the reply, Christoph!


(system) #4